3 ms·

> The drone, which can be directed to detonate on impact, was operating in a "highly effective" autonomous mode that required no human controller, the New York

by setr 5y ago

> The drone, which can be directed to detonate on impact, was operating in a "highly effective" autonomous mode that required no human controller, the New York Post said. > "The lethal autonomous weapons systems were programmed to attack targets without requiring data connectivity between the operator and the munition: in effect, a true 'fire, forget and find' capability," the report from the UN Security Council's Panel of Experts on Libya said. I’m not sure why this is called “rogue” — apparently it was operating as intended

ergot_vacation 5y ago
Probably "rogue" in the sense that direct control was not maintained at all times. The human controller told the drone to pull the trigger when the time was right, and the drone itself made the call as to when that was. It is a bit of sensationalizing for the sake of the headline though.
- setr 5y ago
  the problem is that "rogue" implies some kind of malfunction -- it acted against a stand-down order and killed anyways. In this case, it seems the drone was supposed to do exactly that, and that's perhaps far more dangerous -- this is the first kill where the drone was approved to determine kill targets, and execute that decision, with no human to verify. This really shouldn't be a goal we have in the first place -- to remove humans from the decision loop -- and yet apparently here it is.
  - nradov 5y ago
    This is nothing new. The US military developed a naval mine that could launch a torpedo to kill targets without humans in the decision loop back in 1979. https://en.wikipedia.org/wiki/Mark_60_CAPTOR?wprov=sfla1 https://en.wikipedia.org/wiki/Mark_60_CAPTOR?wprov=sfla1
    - setr 5y ago
      Hm. That’s interesting — it’s an upgraded mine. I think the indiscriminate nature of it makes it more acceptable, at least to me; you’re creating a no-go zone for all — allies, enemies and nuetrals. With this you’re creating a no-go zone for your enemies, and very likely safe for your allies (who you can tag properly with markers), but the difference between enemy and neutral is rather unclear (especially in the Middle East, where the line between combatant and civilian is very thin, almost indistinguishable even to humans on-site)
      - wearywanderer 5y ago
        CAPTOR mines are discriminate. They can recognize different ships by sound and won't fire on friendly ships.
  - PeterisP 5y ago
    I believe the goal is driven by the desire for the drone to continue to operate even if humans are not reachable because of communications jamming - if you're fighting against some low-resourced guerillas, you can have humans in the decision loop, but if the loitering munitions/drones are to be used against an opponent with some air defence and EW/ECM systems (e.g. the recent Azerbaijan-Armenia war, where Turkish drones/loitering munitions systems were widely used, especially against AD systems) then an autonomous operation capability seems a quite desirable goal for practical reasons.
dang 5y ago
This comment was posted when the headline said "rogue" because the URL was https://www.businessinsider.com/killer-drone-hunted-down-human-target-without-being-told-un-2021-5 https://www.businessinsider.com/killer-drone-hunted-down-hum.... We've since changed both (see https://news.ycombinator.com/item?id=27337912 https://news.ycombinator.com/item?id=27337912).
swiley 5y ago
Meh, it's an iteration on guided missiles.
- narrator 5y ago
  Guided missiles have to be told by a human what to strike. The dangerous thing that's happening here is that we trust the AI to kill someone all on its own without the operator having to identify the target.
  - nradov 5y ago
    Not really. There have been anti-ship and anti-armor guided missiles available for years which fly to a designated point and then independently search for a target to attack without a human operator. This is nothing new.
  - tolbish 5y ago
    Iron Dome requires humans to manually target each incoming rocket?
avereveard 5y ago
Also from the description basic ir missiles would apply

by setr 5y ago

ergot_vacation 5y ago

Probably "rogue" in the sense that direct control was not maintained at all times. The human controller told the drone to pull the trigger when the time was right, and the drone itself made the call as to when that was. It is a bit of sensationalizing for the sake of the headline though.

setr 5y ago
the problem is that "rogue" implies some kind of malfunction -- it acted against a stand-down order and killed anyways. In this case, it seems the drone was supposed to do exactly that, and that's perhaps far more dangerous -- this is the first kill where the drone was approved to determine kill targets, and execute that decision, with no human to verify. This really shouldn't be a goal we have in the first place -- to remove humans from the decision loop -- and yet apparently here it is.
- nradov 5y ago
  This is nothing new. The US military developed a naval mine that could launch a torpedo to kill targets without humans in the decision loop back in 1979. https://en.wikipedia.org/wiki/Mark_60_CAPTOR?wprov=sfla1 https://en.wikipedia.org/wiki/Mark_60_CAPTOR?wprov=sfla1
  - setr 5y ago
    Hm. That’s interesting — it’s an upgraded mine. I think the indiscriminate nature of it makes it more acceptable, at least to me; you’re creating a no-go zone for all — allies, enemies and nuetrals. With this you’re creating a no-go zone for your enemies, and very likely safe for your allies (who you can tag properly with markers), but the difference between enemy and neutral is rather unclear (especially in the Middle East, where the line between combatant and civilian is very thin, almost indistinguishable even to humans on-site)
    - wearywanderer 5y ago
      CAPTOR mines are discriminate. They can recognize different ships by sound and won't fire on friendly ships.
- PeterisP 5y ago
  I believe the goal is driven by the desire for the drone to continue to operate even if humans are not reachable because of communications jamming - if you're fighting against some low-resourced guerillas, you can have humans in the decision loop, but if the loitering munitions/drones are to be used against an opponent with some air defence and EW/ECM systems (e.g. the recent Azerbaijan-Armenia war, where Turkish drones/loitering munitions systems were widely used, especially against AD systems) then an autonomous operation capability seems a quite desirable goal for practical reasons.

dang 5y ago

This comment was posted when the headline said "rogue" because the URL was https://www.businessinsider.com/killer-drone-hunted-down-human-target-without-being-told-un-2021-5 https://www.businessinsider.com/killer-drone-hunted-down-hum.... We've since changed both (see https://news.ycombinator.com/item?id=27337912 https://news.ycombinator.com/item?id=27337912).

swiley 5y ago

Meh, it's an iteration on guided missiles.

narrator 5y ago
Guided missiles have to be told by a human what to strike. The dangerous thing that's happening here is that we trust the AI to kill someone all on its own without the operator having to identify the target.
- nradov 5y ago
  Not really. There have been anti-ship and anti-armor guided missiles available for years which fly to a designated point and then independently search for a target to attack without a human operator. This is nothing new.
- tolbish 5y ago
  Iron Dome requires humans to manually target each incoming rocket?

avereveard 5y ago

Also from the description basic ir missiles would apply